A no-code wizard that scaffolds a personal Agent OS for any discipline. A short interview in; a personalized bootstrap prompt for Claude Code out.
The live builder, embedded · scroll inside to try it
The agent OS works because of its structure: named specialists, contracts, rituals, memory. But standing one up meant reading a long bootstrap prompt and knowing what to answer. The builder turns that into a short, human interview, so a designer, a PM, or an analyst can scaffold their own OS without touching a config file.
Four interview screens: you and your discipline, what you ship (your verticals plus the cross-functional lens that pressure-tests everything), your house rules, and your tools and rituals. Then a review gate that reflects your answers back, with your starting team named and the file tree it will propose. The output is a personalized bootstrap prompt: paste it into Claude Code as your first message and it builds the OS, stopping for your approval at every gate.
The builder practices the OS's first principle: it never assumes your role, stack, or standards. It asks, reflects the spec back at a gate, and only then generates.
"Never fabricate" sits at the top of the house-rules step and cannot be unchecked. Every OS the builder produces inherits it as the un-waivable first rule.
The wizard ends in one personalized bootstrap prompt with your Phase 1 answers pre-filled. The OS itself stays plain Markdown in a folder, owned by you, editable by anyone.
The generated prompt scaffolds standards first, agents second, and stops for explicit approval at every phase. The discipline of the system travels with every copy of it.
A working tool, embedded above, generating a complete bootstrap prompt from your answers. It's how the agent OS stops being mine and becomes anyone's.